Background: The inter-rater reliability of the 17-item Hamilton Depression Rating Scale (HDRS) has not been examined in primary care settings with a view to using it as both a primary outcome measure and for determining entry into a trial.
Methods: A semi-structured interview for the HDRS with detailed questions and scoring rules was developed for use by lay interviewers. Interviews with patients with a range of scores on the HDRS were conducted by one interviewer, audiotaped and rated by a second interviewer. 84 ratings were made by 4 raters on 42 primary care patients over 4 sessions, giving comparisons for 10 rater pairs.
Results: The intraclass coefficient and concordance coefficient for the total HDRS score were both 0.95, with a 95% reference interval for the difference (between any pair of raters) in total HDRS score of -3.31 to 3.69; measurement error in an individual total score was 1.25. Inter-rater reliability did not vary with severity of depression and there was no evidence of bias in rating in any rater compared to the others. Weighted kappa coefficients for individual items were close to 0.6 (good) or above for all items except hypochondriasis and insight.
Limitations: The inter-rater reliability of the retardation and agitation items was not assessed.
Conclusions: In primary care, the 17-item HDRS delivered using a standardised interview has high overall inter-rater reliability as a primary outcome measure but a few patients may be erroneously excluded if it is used to determine study entry.